| Line | Branch | Exec | Source |
|---|---|---|---|
| 1 | /* | ||
| 2 | * (c) 2001 Fabrice Bellard | ||
| 3 | * 2007 Marc Hoffman <marc.hoffman@analog.com> | ||
| 4 | * | ||
| 5 | * This file is part of FFmpeg. | ||
| 6 | * | ||
| 7 | * FFmpeg is free software; you can redistribute it and/or | ||
| 8 | * modify it under the terms of the GNU Lesser General Public | ||
| 9 | * License as published by the Free Software Foundation; either | ||
| 10 | * version 2.1 of the License, or (at your option) any later version. | ||
| 11 | * | ||
| 12 | * FFmpeg is distributed in the hope that it will be useful, | ||
| 13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | ||
| 14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | ||
| 15 | * Lesser General Public License for more details. | ||
| 16 | * | ||
| 17 | * You should have received a copy of the GNU Lesser General Public | ||
| 18 | * License along with FFmpeg; if not, write to the Free Software | ||
| 19 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA | ||
| 20 | */ | ||
| 21 | |||
| 22 | /** | ||
| 23 | * @file | ||
| 24 | * DCT test (c) 2001 Fabrice Bellard | ||
| 25 | * Started from sample code by Juan J. Sierralta P. | ||
| 26 | */ | ||
| 27 | |||
| 28 | #include "config.h" | ||
| 29 | #include "config_components.h" | ||
| 30 | #include <stdlib.h> | ||
| 31 | #include <stdio.h> | ||
| 32 | #include <string.h> | ||
| 33 | #if HAVE_UNISTD_H | ||
| 34 | #include <unistd.h> | ||
| 35 | #endif | ||
| 36 | #include <math.h> | ||
| 37 | |||
| 38 | #include "libavutil/cpu.h" | ||
| 39 | #include "libavutil/common.h" | ||
| 40 | #include "libavutil/internal.h" | ||
| 41 | #include "libavutil/lfg.h" | ||
| 42 | #include "libavutil/mem_internal.h" | ||
| 43 | #include "libavutil/time.h" | ||
| 44 | |||
| 45 | #include "libavcodec/dct.h" | ||
| 46 | #include "libavcodec/fdctdsp.h" | ||
| 47 | #include "libavcodec/idctdsp.h" | ||
| 48 | #include "libavcodec/simple_idct.h" | ||
| 49 | #include "libavcodec/xvididct.h" | ||
| 50 | #include "libavcodec/aandcttab.h" | ||
| 51 | #include "libavcodec/faandct.h" | ||
| 52 | #include "libavcodec/faanidct.h" | ||
| 53 | #include "libavcodec/dctref.h" | ||
| 54 | #if CONFIG_PRORES_DECODER | ||
| 55 | #include "libavcodec/proresdsp.c" | ||
| 56 | #endif | ||
| 57 | |||
| 58 | struct algo { | ||
| 59 | const char *name; | ||
| 60 | void (*func)(int16_t *block); | ||
| 61 | enum idct_permutation_type perm_type; | ||
| 62 | int cpu_flag; | ||
| 63 | int nonspec; | ||
| 64 | }; | ||
| 65 | |||
| 66 | static const struct algo fdct_tab[] = { | ||
| 67 | { "REF-DBL", ff_ref_fdct, FF_IDCT_PERM_NONE }, | ||
| 68 | { "IJG-AAN-INT", ff_fdct_ifast, FF_IDCT_PERM_NONE }, | ||
| 69 | { "IJG-LLM-INT", ff_jpeg_fdct_islow_8, FF_IDCT_PERM_NONE }, | ||
| 70 | #if CONFIG_FAANDCT | ||
| 71 | { "FAAN", ff_faandct, FF_IDCT_PERM_NONE }, | ||
| 72 | #endif /* CONFIG_FAANDCT */ | ||
| 73 | }; | ||
| 74 | |||
| 75 | #if CONFIG_PRORES_DECODER | ||
| 76 | 60000 | static void ff_prores_idct_wrap(int16_t *dst){ | |
| 77 | 60000 | LOCAL_ALIGNED(16, int16_t, qmat, [64]); | |
| 78 | int i; | ||
| 79 | |||
| 80 |
2/2✓ Branch 0 taken 3840000 times.
✓ Branch 1 taken 60000 times.
|
3900000 | for(i=0; i<64; i++){ |
| 81 | 3840000 | qmat[i]=4; | |
| 82 | } | ||
| 83 | 60000 | prores_idct_10(dst, qmat); | |
| 84 |
2/2✓ Branch 0 taken 3840000 times.
✓ Branch 1 taken 60000 times.
|
3900000 | for(i=0; i<64; i++) { |
| 85 | 3840000 | dst[i] -= 512; | |
| 86 | } | ||
| 87 | 60000 | } | |
| 88 | #endif | ||
| 89 | |||
| 90 | static const struct algo idct_tab[] = { | ||
| 91 | { "REF-DBL", ff_ref_idct, FF_IDCT_PERM_NONE }, | ||
| 92 | { "INT", ff_j_rev_dct, FF_IDCT_PERM_LIBMPEG2 }, | ||
| 93 | { "SIMPLE-C", ff_simple_idct_int16_8bit, FF_IDCT_PERM_NONE }, | ||
| 94 | { "SIMPLE-C10", ff_simple_idct_int16_10bit, FF_IDCT_PERM_NONE }, | ||
| 95 | { "SIMPLE-C12", ff_simple_idct_int16_12bit, FF_IDCT_PERM_NONE, 0, 1 }, | ||
| 96 | #if CONFIG_PRORES_DECODER | ||
| 97 | { "PR-C", ff_prores_idct_wrap, FF_IDCT_PERM_NONE, 0, 1 }, | ||
| 98 | #endif | ||
| 99 | #if CONFIG_FAANIDCT | ||
| 100 | { "FAANI", ff_faanidct, FF_IDCT_PERM_NONE }, | ||
| 101 | #endif /* CONFIG_FAANIDCT */ | ||
| 102 | #if CONFIG_MPEG4_DECODER | ||
| 103 | { "XVID", ff_xvid_idct, FF_IDCT_PERM_NONE, 0, 1 }, | ||
| 104 | #endif /* CONFIG_MPEG4_DECODER */ | ||
| 105 | }; | ||
| 106 | |||
| 107 | #if ARCH_AARCH64 | ||
| 108 | #include "aarch64/dct.c" | ||
| 109 | #elif ARCH_ARM | ||
| 110 | #include "arm/dct.c" | ||
| 111 | #elif ARCH_PPC | ||
| 112 | #include "ppc/dct.c" | ||
| 113 | #elif ARCH_X86 | ||
| 114 | #include "x86/dct.c" | ||
| 115 | #else | ||
| 116 | static const struct algo fdct_tab_arch[] = { { 0 } }; | ||
| 117 | static const struct algo idct_tab_arch[] = { { 0 } }; | ||
| 118 | #endif | ||
| 119 | |||
| 120 | #define AANSCALE_BITS 12 | ||
| 121 | |||
| 122 | #define NB_ITS 20000 | ||
| 123 | #define NB_ITS_SPEED 50000 | ||
| 124 | |||
| 125 | DECLARE_ALIGNED(16, static int16_t, block)[64]; | ||
| 126 | DECLARE_ALIGNED(8, static int16_t, block1)[64]; | ||
| 127 | |||
| 128 | 1120000 | static void init_block(int16_t block[64], int test, int is_idct, AVLFG *prng, int vals) | |
| 129 | { | ||
| 130 | int i, j; | ||
| 131 | |||
| 132 | 1120000 | memset(block, 0, 64 * sizeof(*block)); | |
| 133 | |||
| 134 |
3/4✓ Branch 0 taken 340000 times.
✓ Branch 1 taken 440000 times.
✓ Branch 2 taken 340000 times.
✗ Branch 3 not taken.
|
1120000 | switch (test) { |
| 135 | 340000 | case 0: | |
| 136 |
2/2✓ Branch 0 taken 21760000 times.
✓ Branch 1 taken 340000 times.
|
22100000 | for (i = 0; i < 64; i++) |
| 137 | 21760000 | block[i] = (av_lfg_get(prng) % (2*vals)) -vals; | |
| 138 |
1/2✓ Branch 0 taken 340000 times.
✗ Branch 1 not taken.
|
340000 | if (is_idct) { |
| 139 | 340000 | ff_ref_fdct(block); | |
| 140 |
2/2✓ Branch 0 taken 21760000 times.
✓ Branch 1 taken 340000 times.
|
22100000 | for (i = 0; i < 64; i++) |
| 141 | 21760000 | block[i] >>= 3; | |
| 142 | } | ||
| 143 | 340000 | break; | |
| 144 | 440000 | case 1: | |
| 145 | 440000 | j = av_lfg_get(prng) % 10 + 1; | |
| 146 |
2/2✓ Branch 0 taken 2413686 times.
✓ Branch 1 taken 440000 times.
|
2853686 | for (i = 0; i < j; i++) { |
| 147 | 2413686 | int idx = av_lfg_get(prng) % 64; | |
| 148 | 2413686 | block[idx] = av_lfg_get(prng) % (2*vals) -vals; | |
| 149 | } | ||
| 150 | 440000 | break; | |
| 151 | 340000 | case 2: | |
| 152 | 340000 | block[ 0] = av_lfg_get(prng) % (16*vals) - (8*vals); | |
| 153 | 340000 | block[63] = (block[0] & 1) ^ 1; | |
| 154 | 340000 | break; | |
| 155 | } | ||
| 156 | 1120000 | } | |
| 157 | |||
| 158 | 1120000 | static void permute(int16_t dst[64], const int16_t src[64], | |
| 159 | enum idct_permutation_type perm_type) | ||
| 160 | { | ||
| 161 | int i; | ||
| 162 | |||
| 163 | #if ARCH_X86 | ||
| 164 |
2/2✓ Branch 1 taken 60000 times.
✓ Branch 2 taken 1060000 times.
|
1120000 | if (permute_x86(dst, src, perm_type)) |
| 165 | 60000 | return; | |
| 166 | #endif | ||
| 167 | |||
| 168 |
3/4✓ Branch 0 taken 60000 times.
✗ Branch 1 not taken.
✓ Branch 2 taken 480000 times.
✓ Branch 3 taken 520000 times.
|
1060000 | switch (perm_type) { |
| 169 | 60000 | case FF_IDCT_PERM_LIBMPEG2: | |
| 170 |
2/2✓ Branch 0 taken 3840000 times.
✓ Branch 1 taken 60000 times.
|
3900000 | for (i = 0; i < 64; i++) |
| 171 | 3840000 | dst[(i & 0x38) | ((i & 6) >> 1) | ((i & 1) << 2)] = src[i]; | |
| 172 | 60000 | break; | |
| 173 | ✗ | case FF_IDCT_PERM_PARTTRANS: | |
| 174 | ✗ | for (i = 0; i < 64; i++) | |
| 175 | ✗ | dst[(i & 0x24) | ((i & 3) << 3) | ((i >> 3) & 3)] = src[i]; | |
| 176 | ✗ | break; | |
| 177 | 480000 | case FF_IDCT_PERM_TRANSPOSE: | |
| 178 |
2/2✓ Branch 0 taken 30720000 times.
✓ Branch 1 taken 480000 times.
|
31200000 | for (i = 0; i < 64; i++) |
| 179 | 30720000 | dst[(i>>3) | ((i<<3)&0x38)] = src[i]; | |
| 180 | 480000 | break; | |
| 181 | 520000 | default: | |
| 182 |
2/2✓ Branch 0 taken 33280000 times.
✓ Branch 1 taken 520000 times.
|
33800000 | for (i = 0; i < 64; i++) |
| 183 | 33280000 | dst[i] = src[i]; | |
| 184 | 520000 | break; | |
| 185 | } | ||
| 186 | } | ||
| 187 | |||
| 188 | 56 | static int dct_error(const struct algo *dct, int test, int is_idct, int speed, const int bits) | |
| 189 | { | ||
| 190 |
2/2✓ Branch 0 taken 51 times.
✓ Branch 1 taken 5 times.
|
56 | void (*ref)(int16_t *block) = is_idct ? ff_ref_idct : ff_ref_fdct; |
| 191 | int it, i, scale; | ||
| 192 | int err_inf, v; | ||
| 193 | 56 | int64_t err2, ti, ti1, it1, err_sum = 0; | |
| 194 | 56 | int64_t sysErr[64], sysErrMax = 0; | |
| 195 | 56 | int64_t err2_matrix[64], err2_max = 0; | |
| 196 | 56 | int maxout = 0; | |
| 197 | 56 | int blockSumErrMax = 0, blockSumErr; | |
| 198 | AVLFG prng; | ||
| 199 | 56 | const int vals=1<<bits; | |
| 200 | double omse, ome; | ||
| 201 | int spec_err; | ||
| 202 | |||
| 203 | 56 | av_lfg_init(&prng, 1); | |
| 204 | |||
| 205 | 56 | err_inf = 0; | |
| 206 | 56 | err2 = 0; | |
| 207 |
2/2✓ Branch 0 taken 3584 times.
✓ Branch 1 taken 56 times.
|
3640 | for (i = 0; i < 64; i++) |
| 208 | 3584 | err2_matrix[i] = sysErr[i] = 0; | |
| 209 |
2/2✓ Branch 0 taken 1120000 times.
✓ Branch 1 taken 56 times.
|
1120056 | for (it = 0; it < NB_ITS; it++) { |
| 210 | 1120000 | init_block(block1, test, is_idct, &prng, vals); | |
| 211 | 1120000 | permute(block, block1, dct->perm_type); | |
| 212 | |||
| 213 | 1120000 | dct->func(block); | |
| 214 | |||
| 215 |
2/2✓ Branch 0 taken 20000 times.
✓ Branch 1 taken 1100000 times.
|
1120000 | if (!strcmp(dct->name, "IJG-AAN-INT")) { |
| 216 |
2/2✓ Branch 0 taken 1280000 times.
✓ Branch 1 taken 20000 times.
|
1300000 | for (i = 0; i < 64; i++) { |
| 217 | 1280000 | scale = 8 * (1 << (AANSCALE_BITS + 11)) / ff_aanscales[i]; | |
| 218 | 1280000 | block[i] = (block[i] * scale) >> AANSCALE_BITS; | |
| 219 | } | ||
| 220 | } | ||
| 221 | |||
| 222 | 1120000 | ref(block1); | |
| 223 |
2/2✓ Branch 0 taken 60000 times.
✓ Branch 1 taken 1060000 times.
|
1120000 | if (!strcmp(dct->name, "PR-SSE2")) |
| 224 |
2/2✓ Branch 0 taken 3840000 times.
✓ Branch 1 taken 60000 times.
|
3900000 | for (i = 0; i < 64; i++) |
| 225 | 3840000 | block1[i] = av_clip(block1[i], 4-512, 1019-512); | |
| 226 | |||
| 227 | 1120000 | blockSumErr = 0; | |
| 228 |
2/2✓ Branch 0 taken 71680000 times.
✓ Branch 1 taken 1120000 times.
|
72800000 | for (i = 0; i < 64; i++) { |
| 229 | 71680000 | int err = block[i] - block1[i]; | |
| 230 | 71680000 | err_sum += err; | |
| 231 | 71680000 | v = abs(err); | |
| 232 |
2/2✓ Branch 0 taken 61 times.
✓ Branch 1 taken 71679939 times.
|
71680000 | if (v > err_inf) |
| 233 | 61 | err_inf = v; | |
| 234 | 71680000 | err2_matrix[i] += v * (int64_t)v; | |
| 235 | 71680000 | err2 += v * (int64_t)v; | |
| 236 | 71680000 | sysErr[i] += block[i] - block1[i]; | |
| 237 | 71680000 | blockSumErr += v; | |
| 238 |
2/2✓ Branch 0 taken 756 times.
✓ Branch 1 taken 71679244 times.
|
71680000 | if (abs(block[i]) > maxout) |
| 239 | 756 | maxout = abs(block[i]); | |
| 240 | } | ||
| 241 |
2/2✓ Branch 0 taken 246 times.
✓ Branch 1 taken 1119754 times.
|
1120000 | if (blockSumErrMax < blockSumErr) |
| 242 | 246 | blockSumErrMax = blockSumErr; | |
| 243 | } | ||
| 244 |
2/2✓ Branch 0 taken 3584 times.
✓ Branch 1 taken 56 times.
|
3640 | for (i = 0; i < 64; i++) { |
| 245 | 3584 | sysErrMax = FFMAX(sysErrMax, FFABS(sysErr[i])); | |
| 246 | 3584 | err2_max = FFMAX(err2_max , FFABS(err2_matrix[i])); | |
| 247 | } | ||
| 248 | |||
| 249 |
2/2✓ Branch 0 taken 3584 times.
✓ Branch 1 taken 56 times.
|
3640 | for (i = 0; i < 64; i++) { |
| 250 |
2/2✓ Branch 0 taken 448 times.
✓ Branch 1 taken 3136 times.
|
3584 | if (i % 8 == 0) |
| 251 | 448 | printf("\n"); | |
| 252 | 3584 | printf("%7d ", (int) sysErr[i]); | |
| 253 | } | ||
| 254 | 56 | printf("\n"); | |
| 255 | |||
| 256 | 56 | omse = (double) err2 / NB_ITS / 64; | |
| 257 | 56 | ome = (double) err_sum / NB_ITS / 64; | |
| 258 | |||
| 259 |
7/8✓ Branch 0 taken 51 times.
✓ Branch 1 taken 5 times.
✓ Branch 2 taken 51 times.
✗ Branch 3 not taken.
✓ Branch 4 taken 36 times.
✓ Branch 5 taken 15 times.
✓ Branch 6 taken 4 times.
✓ Branch 7 taken 32 times.
|
56 | spec_err = is_idct && (err_inf > 1 || omse > 0.02 || fabs(ome) > 0.0015); |
| 260 |
2/2✓ Branch 0 taken 39 times.
✓ Branch 1 taken 17 times.
|
56 | if (test < 2) |
| 261 |
6/6✓ Branch 0 taken 34 times.
✓ Branch 1 taken 5 times.
✓ Branch 2 taken 28 times.
✓ Branch 3 taken 6 times.
✓ Branch 4 taken 2 times.
✓ Branch 5 taken 26 times.
|
39 | spec_err = is_idct && ((double) err2_max / NB_ITS > 0.06 || (double) sysErrMax / NB_ITS > 0.015); |
| 262 | |||
| 263 | 56 | printf("%s %s: max_err=%d omse=%0.8f ome=%0.8f syserr=%0.8f maxout=%d blockSumErr=%d\n", | |
| 264 | 56 | is_idct ? "IDCT" : "DCT", dct->name, err_inf, | |
| 265 |
2/2✓ Branch 0 taken 51 times.
✓ Branch 1 taken 5 times.
|
56 | omse, ome, (double) sysErrMax / NB_ITS, |
| 266 | maxout, blockSumErrMax); | ||
| 267 | |||
| 268 |
3/4✓ Branch 0 taken 13 times.
✓ Branch 1 taken 43 times.
✗ Branch 2 not taken.
✓ Branch 3 taken 13 times.
|
56 | if (spec_err && !dct->nonspec) { |
| 269 | ✗ | printf("Failed!\n"); | |
| 270 | ✗ | return 1; | |
| 271 | } | ||
| 272 | |||
| 273 |
1/2✓ Branch 0 taken 56 times.
✗ Branch 1 not taken.
|
56 | if (!speed) |
| 274 | 56 | return 0; | |
| 275 | |||
| 276 | /* speed test */ | ||
| 277 | |||
| 278 | ✗ | init_block(block, test, is_idct, &prng, vals); | |
| 279 | ✗ | permute(block1, block, dct->perm_type); | |
| 280 | |||
| 281 | ✗ | ti = av_gettime_relative(); | |
| 282 | ✗ | it1 = 0; | |
| 283 | do { | ||
| 284 | ✗ | for (it = 0; it < NB_ITS_SPEED; it++) { | |
| 285 | ✗ | memcpy(block, block1, sizeof(block)); | |
| 286 | ✗ | dct->func(block); | |
| 287 | } | ||
| 288 | ✗ | it1 += NB_ITS_SPEED; | |
| 289 | ✗ | ti1 = av_gettime_relative() - ti; | |
| 290 | ✗ | } while (ti1 < 1000000); | |
| 291 | |||
| 292 | ✗ | printf("%s %s: %0.1f kdct/s\n", is_idct ? "IDCT" : "DCT", dct->name, | |
| 293 | ✗ | (double) it1 * 1000.0 / (double) ti1); | |
| 294 | |||
| 295 | ✗ | return 0; | |
| 296 | } | ||
| 297 | |||
| 298 | DECLARE_ALIGNED(8, static uint8_t, img_dest)[64]; | ||
| 299 | DECLARE_ALIGNED(8, static uint8_t, img_dest1)[64]; | ||
| 300 | |||
| 301 | 20000 | static void idct248_ref(uint8_t *dest, ptrdiff_t linesize, int16_t *block) | |
| 302 | { | ||
| 303 | static int init; | ||
| 304 | static double c8[8][8]; | ||
| 305 | static double c4[4][4]; | ||
| 306 | double block1[64], block2[64], block3[64]; | ||
| 307 | double s, sum, v; | ||
| 308 | int i, j, k; | ||
| 309 | |||
| 310 |
2/2✓ Branch 0 taken 1 times.
✓ Branch 1 taken 19999 times.
|
20000 | if (!init) { |
| 311 | 1 | init = 1; | |
| 312 | |||
| 313 |
2/2✓ Branch 0 taken 8 times.
✓ Branch 1 taken 1 times.
|
9 | for (i = 0; i < 8; i++) { |
| 314 | 8 | sum = 0; | |
| 315 |
2/2✓ Branch 0 taken 64 times.
✓ Branch 1 taken 8 times.
|
72 | for (j = 0; j < 8; j++) { |
| 316 |
2/2✓ Branch 0 taken 8 times.
✓ Branch 1 taken 56 times.
|
64 | s = (i == 0) ? sqrt(1.0 / 8.0) : sqrt(1.0 / 4.0); |
| 317 | 64 | c8[i][j] = s * cos(M_PI * i * (j + 0.5) / 8.0); | |
| 318 | 64 | sum += c8[i][j] * c8[i][j]; | |
| 319 | } | ||
| 320 | } | ||
| 321 | |||
| 322 |
2/2✓ Branch 0 taken 4 times.
✓ Branch 1 taken 1 times.
|
5 | for (i = 0; i < 4; i++) { |
| 323 | 4 | sum = 0; | |
| 324 |
2/2✓ Branch 0 taken 16 times.
✓ Branch 1 taken 4 times.
|
20 | for (j = 0; j < 4; j++) { |
| 325 |
2/2✓ Branch 0 taken 4 times.
✓ Branch 1 taken 12 times.
|
16 | s = (i == 0) ? sqrt(1.0 / 4.0) : sqrt(1.0 / 2.0); |
| 326 | 16 | c4[i][j] = s * cos(M_PI * i * (j + 0.5) / 4.0); | |
| 327 | 16 | sum += c4[i][j] * c4[i][j]; | |
| 328 | } | ||
| 329 | } | ||
| 330 | } | ||
| 331 | |||
| 332 | /* butterfly */ | ||
| 333 | 20000 | s = 0.5 * sqrt(2.0); | |
| 334 |
2/2✓ Branch 0 taken 80000 times.
✓ Branch 1 taken 20000 times.
|
100000 | for (i = 0; i < 4; i++) { |
| 335 |
2/2✓ Branch 0 taken 640000 times.
✓ Branch 1 taken 80000 times.
|
720000 | for (j = 0; j < 8; j++) { |
| 336 | 640000 | block1[8 * (2 * i) + j] = | |
| 337 | 640000 | (block[8 * (2 * i) + j] + block[8 * (2 * i + 1) + j]) * s; | |
| 338 | 640000 | block1[8 * (2 * i + 1) + j] = | |
| 339 | 640000 | (block[8 * (2 * i) + j] - block[8 * (2 * i + 1) + j]) * s; | |
| 340 | } | ||
| 341 | } | ||
| 342 | |||
| 343 | /* idct8 on lines */ | ||
| 344 |
2/2✓ Branch 0 taken 160000 times.
✓ Branch 1 taken 20000 times.
|
180000 | for (i = 0; i < 8; i++) { |
| 345 |
2/2✓ Branch 0 taken 1280000 times.
✓ Branch 1 taken 160000 times.
|
1440000 | for (j = 0; j < 8; j++) { |
| 346 | 1280000 | sum = 0; | |
| 347 |
2/2✓ Branch 0 taken 10240000 times.
✓ Branch 1 taken 1280000 times.
|
11520000 | for (k = 0; k < 8; k++) |
| 348 | 10240000 | sum += c8[k][j] * block1[8 * i + k]; | |
| 349 | 1280000 | block2[8 * i + j] = sum; | |
| 350 | } | ||
| 351 | } | ||
| 352 | |||
| 353 | /* idct4 */ | ||
| 354 |
2/2✓ Branch 0 taken 160000 times.
✓ Branch 1 taken 20000 times.
|
180000 | for (i = 0; i < 8; i++) { |
| 355 |
2/2✓ Branch 0 taken 640000 times.
✓ Branch 1 taken 160000 times.
|
800000 | for (j = 0; j < 4; j++) { |
| 356 | /* top */ | ||
| 357 | 640000 | sum = 0; | |
| 358 |
2/2✓ Branch 0 taken 2560000 times.
✓ Branch 1 taken 640000 times.
|
3200000 | for (k = 0; k < 4; k++) |
| 359 | 2560000 | sum += c4[k][j] * block2[8 * (2 * k) + i]; | |
| 360 | 640000 | block3[8 * (2 * j) + i] = sum; | |
| 361 | |||
| 362 | /* bottom */ | ||
| 363 | 640000 | sum = 0; | |
| 364 |
2/2✓ Branch 0 taken 2560000 times.
✓ Branch 1 taken 640000 times.
|
3200000 | for (k = 0; k < 4; k++) |
| 365 | 2560000 | sum += c4[k][j] * block2[8 * (2 * k + 1) + i]; | |
| 366 | 640000 | block3[8 * (2 * j + 1) + i] = sum; | |
| 367 | } | ||
| 368 | } | ||
| 369 | |||
| 370 | /* clamp and store the result */ | ||
| 371 |
2/2✓ Branch 0 taken 160000 times.
✓ Branch 1 taken 20000 times.
|
180000 | for (i = 0; i < 8; i++) { |
| 372 |
2/2✓ Branch 0 taken 1280000 times.
✓ Branch 1 taken 160000 times.
|
1440000 | for (j = 0; j < 8; j++) { |
| 373 | 1280000 | v = block3[8 * i + j]; | |
| 374 |
2/2✓ Branch 0 taken 53473 times.
✓ Branch 1 taken 1226527 times.
|
1280000 | if (v < 0) v = 0; |
| 375 |
2/2✓ Branch 0 taken 54705 times.
✓ Branch 1 taken 1171822 times.
|
1226527 | else if (v > 255) v = 255; |
| 376 | 1280000 | dest[i * linesize + j] = (int) rint(v); | |
| 377 | } | ||
| 378 | } | ||
| 379 | 20000 | } | |
| 380 | |||
| 381 | 1 | static void idct248_error(const char *name, | |
| 382 | void (*idct248_put)(uint8_t *dest, | ||
| 383 | ptrdiff_t line_size, | ||
| 384 | int16_t *block), | ||
| 385 | int speed) | ||
| 386 | { | ||
| 387 | int it, i, it1, ti, ti1, err_max, v; | ||
| 388 | AVLFG prng; | ||
| 389 | |||
| 390 | 1 | av_lfg_init(&prng, 1); | |
| 391 | |||
| 392 | /* just one test to see if code is correct (precision is less | ||
| 393 | important here) */ | ||
| 394 | 1 | err_max = 0; | |
| 395 |
2/2✓ Branch 0 taken 20000 times.
✓ Branch 1 taken 1 times.
|
20001 | for (it = 0; it < NB_ITS; it++) { |
| 396 | /* XXX: use forward transform to generate values */ | ||
| 397 |
2/2✓ Branch 0 taken 1280000 times.
✓ Branch 1 taken 20000 times.
|
1300000 | for (i = 0; i < 64; i++) |
| 398 | 1280000 | block1[i] = av_lfg_get(&prng) % 256 - 128; | |
| 399 | 20000 | block1[0] += 1024; | |
| 400 | |||
| 401 |
2/2✓ Branch 0 taken 1280000 times.
✓ Branch 1 taken 20000 times.
|
1300000 | for (i = 0; i < 64; i++) |
| 402 | 1280000 | block[i] = block1[i]; | |
| 403 | 20000 | idct248_ref(img_dest1, 8, block); | |
| 404 | |||
| 405 |
2/2✓ Branch 0 taken 1280000 times.
✓ Branch 1 taken 20000 times.
|
1300000 | for (i = 0; i < 64; i++) |
| 406 | 1280000 | block[i] = block1[i]; | |
| 407 | 20000 | idct248_put(img_dest, 8, block); | |
| 408 | |||
| 409 |
2/2✓ Branch 0 taken 1280000 times.
✓ Branch 1 taken 20000 times.
|
1300000 | for (i = 0; i < 64; i++) { |
| 410 | 1280000 | v = abs((int) img_dest[i] - (int) img_dest1[i]); | |
| 411 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 1280000 times.
|
1280000 | if (v == 255) |
| 412 | ✗ | printf("%d %d\n", img_dest[i], img_dest1[i]); | |
| 413 |
2/2✓ Branch 0 taken 1 times.
✓ Branch 1 taken 1279999 times.
|
1280000 | if (v > err_max) |
| 414 | 1 | err_max = v; | |
| 415 | } | ||
| 416 | #if 0 | ||
| 417 | printf("ref=\n"); | ||
| 418 | for(i=0;i<8;i++) { | ||
| 419 | int j; | ||
| 420 | for(j=0;j<8;j++) { | ||
| 421 | printf(" %3d", img_dest1[i*8+j]); | ||
| 422 | } | ||
| 423 | printf("\n"); | ||
| 424 | } | ||
| 425 | |||
| 426 | printf("out=\n"); | ||
| 427 | for(i=0;i<8;i++) { | ||
| 428 | int j; | ||
| 429 | for(j=0;j<8;j++) { | ||
| 430 | printf(" %3d", img_dest[i*8+j]); | ||
| 431 | } | ||
| 432 | printf("\n"); | ||
| 433 | } | ||
| 434 | #endif | ||
| 435 | } | ||
| 436 | 1 | printf("%s %s: err_inf=%d\n", 1 ? "IDCT248" : "DCT248", name, err_max); | |
| 437 | |||
| 438 |
1/2✓ Branch 0 taken 1 times.
✗ Branch 1 not taken.
|
1 | if (!speed) |
| 439 | 1 | return; | |
| 440 | |||
| 441 | ✗ | ti = av_gettime_relative(); | |
| 442 | ✗ | it1 = 0; | |
| 443 | do { | ||
| 444 | ✗ | for (it = 0; it < NB_ITS_SPEED; it++) { | |
| 445 | ✗ | for (i = 0; i < 64; i++) | |
| 446 | ✗ | block[i] = block1[i]; | |
| 447 | ✗ | idct248_put(img_dest, 8, block); | |
| 448 | } | ||
| 449 | ✗ | it1 += NB_ITS_SPEED; | |
| 450 | ✗ | ti1 = av_gettime_relative() - ti; | |
| 451 | ✗ | } while (ti1 < 1000000); | |
| 452 | |||
| 453 | ✗ | printf("%s %s: %0.1f kdct/s\n", 1 ? "IDCT248" : "DCT248", name, | |
| 454 | ✗ | (double) it1 * 1000.0 / (double) ti1); | |
| 455 | } | ||
| 456 | |||
| 457 | ✗ | static void help(void) | |
| 458 | { | ||
| 459 | ✗ | printf("dct-test [-i] [<test-number>] [<bits>]\n" | |
| 460 | "test-number 0 -> test with random matrixes\n" | ||
| 461 | " 1 -> test with random sparse matrixes\n" | ||
| 462 | " 2 -> do 3. test from MPEG-4 std\n" | ||
| 463 | "bits Number of time domain bits to use, 8 is default\n" | ||
| 464 | "-i test IDCT implementations\n" | ||
| 465 | "-4 test IDCT248 implementations\n" | ||
| 466 | "-t speed test\n"); | ||
| 467 | ✗ | } | |
| 468 | |||
| 469 | #if !HAVE_GETOPT | ||
| 470 | #include "compat/getopt.c" | ||
| 471 | #endif | ||
| 472 | |||
| 473 | 5 | int main(int argc, char **argv) | |
| 474 | { | ||
| 475 | 5 | int test_idct = 0, test_248_dct = 0; | |
| 476 | int c, i; | ||
| 477 | 5 | int test = 1; | |
| 478 | 5 | int speed = 0; | |
| 479 | 5 | int err = 0; | |
| 480 | 5 | int bits=8; | |
| 481 | |||
| 482 | 5 | ff_ref_dct_init(); | |
| 483 | |||
| 484 | for (;;) { | ||
| 485 | 9 | c = getopt(argc, argv, "ih4t"); | |
| 486 |
2/2✓ Branch 0 taken 5 times.
✓ Branch 1 taken 4 times.
|
9 | if (c == -1) |
| 487 | 5 | break; | |
| 488 |
2/4✓ Branch 0 taken 3 times.
✓ Branch 1 taken 1 times.
✗ Branch 2 not taken.
✗ Branch 3 not taken.
|
4 | switch (c) { |
| 489 | 3 | case 'i': | |
| 490 | 3 | test_idct = 1; | |
| 491 | 3 | break; | |
| 492 | 1 | case '4': | |
| 493 | 1 | test_248_dct = 1; | |
| 494 | 1 | break; | |
| 495 | ✗ | case 't': | |
| 496 | ✗ | speed = 1; | |
| 497 | ✗ | break; | |
| 498 | ✗ | default: | |
| 499 | case 'h': | ||
| 500 | ✗ | help(); | |
| 501 | ✗ | return 0; | |
| 502 | } | ||
| 503 | } | ||
| 504 | |||
| 505 |
2/2✓ Branch 0 taken 3 times.
✓ Branch 1 taken 2 times.
|
5 | if (optind < argc) |
| 506 | 3 | test = atoi(argv[optind]); | |
| 507 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 5 times.
|
5 | if(optind+1 < argc) bits= atoi(argv[optind+1]); |
| 508 | |||
| 509 | 5 | printf("ffmpeg DCT/IDCT test\n"); | |
| 510 | |||
| 511 |
2/2✓ Branch 0 taken 1 times.
✓ Branch 1 taken 4 times.
|
5 | if (test_248_dct) { |
| 512 | 1 | idct248_error("SIMPLE-C", ff_simple_idct248_put, speed); | |
| 513 | } else { | ||
| 514 | 4 | const int cpu_flags = av_get_cpu_flags(); | |
| 515 |
2/2✓ Branch 0 taken 3 times.
✓ Branch 1 taken 1 times.
|
4 | if (test_idct) { |
| 516 |
2/2✓ Branch 0 taken 24 times.
✓ Branch 1 taken 3 times.
|
27 | for (i = 0; i < FF_ARRAY_ELEMS(idct_tab); i++) |
| 517 | 24 | err |= dct_error(&idct_tab[i], test, test_idct, speed, bits); | |
| 518 | |||
| 519 |
2/2✓ Branch 0 taken 27 times.
✓ Branch 1 taken 3 times.
|
30 | for (i = 0; idct_tab_arch[i].name; i++) |
| 520 |
1/2✓ Branch 0 taken 27 times.
✗ Branch 1 not taken.
|
27 | if (!(~cpu_flags & idct_tab_arch[i].cpu_flag)) |
| 521 | 27 | err |= dct_error(&idct_tab_arch[i], test, test_idct, speed, bits); | |
| 522 | } | ||
| 523 | #if CONFIG_FDCTDSP | ||
| 524 | else { | ||
| 525 |
2/2✓ Branch 0 taken 4 times.
✓ Branch 1 taken 1 times.
|
5 | for (i = 0; i < FF_ARRAY_ELEMS(fdct_tab); i++) |
| 526 | 4 | err |= dct_error(&fdct_tab[i], test, test_idct, speed, bits); | |
| 527 | |||
| 528 |
2/2✓ Branch 0 taken 1 times.
✓ Branch 1 taken 1 times.
|
2 | for (i = 0; fdct_tab_arch[i].name; i++) |
| 529 |
1/2✓ Branch 0 taken 1 times.
✗ Branch 1 not taken.
|
1 | if (!(~cpu_flags & fdct_tab_arch[i].cpu_flag)) |
| 530 | 1 | err |= dct_error(&fdct_tab_arch[i], test, test_idct, speed, bits); | |
| 531 | } | ||
| 532 | #endif /* CONFIG_FDCTDSP */ | ||
| 533 | } | ||
| 534 | |||
| 535 |
1/2✗ Branch 0 not taken.
✓ Branch 1 taken 5 times.
|
5 | if (err) |
| 536 | ✗ | printf("Error: %d.\n", err); | |
| 537 | |||
| 538 | 5 | return !!err; | |
| 539 | } | ||
| 540 |